This study is an attempt to reduce sub-group differences, while maintaining validity, by using alternative selection devices commonly seen in the public sector and expanding the criterion space by incorporating service-oriented behaviors. Participants were correctional security officers hired between 1999 and 2000 in a Midwestern state. Logistic regression results indicated that it is possible to significantly reduce sub-group differences and have greater validity for a composite of alternative predictors (i.e., Job Knowledge, Interpersonal Skills, Training and Experience) than a composite made up of specific job-related cognitive skills (i.e., Basic Math, Reading Comprehension, Attention to Detail, Memory Recall, Logic, Best English Usage, and Report Writing). Overall, this study provides encouraging results and information for selection system decision-makers in the public sector.